A Third of Perioperative Blood Transfusions in the ICU Does Not Follow Guideline Recommendations-A Retrospective Analysis

Objective: This study evaluates the relevance of transfusion guidelines for clinical transfusion practice. Background: There is little data available on current practice related to appropriate use of blood products. Recent data suggest incorrect use and over transfusion in Europe. In Germany, indications for transfusion and administration of blood products are strict and detailed. However, in clinical practice, alignment of transfusion guidelines to clinical situations seems to be difficult for physiciansespecially in complex diseases and severity such as in critical care. We hypothesized that significant practice variability exists with regard to guidelines adherence in critically ill. Materials and methods: Data sets of transfused patients from a surgical intensive care unit of a university center retrospectively were analyzed over a 12 month period. Indications for blood products were compared with guidelines focusing on numbers of ordered and administered units as well as transfusion triggers. Results: In total, during the study period, 450 packed red cells (PRC, 249 orders), 454 fresh frozen plasma (FFP) (201) and 43 platelet units (PC) (29) were given to 89 patients. The mean number of administered PRC was 5.8+/-6.6 (mean +/SD) units. Double units were administered in 57.4%. Discordant to actual guidelines, 75 (30.12%) PRC`s, 79 (39.30%) FFP's and 9 (31.03%) of PC orders were given without indication. Fresh frozen plasma was administered in ineffective dosage in 83.6% (mean 7.3 ml/kg body weight), platelet transfusion despite intact platelet function in 30.2% and beyond platelet counts of 100 000/μl in 60%. Conclusion: Detailed guidelines for transfusion of blood products were not followed in daily practice in a third of all applications. The reasons for malcompliance are key for improvement.


Introduction
Appropriate use of blood products is one of the "pillars" identified by the Patient Blood Management approach to reduce the use of blood transfusion and to improve the patient outcome. However, various studies demonstrate that inappropriate use of blood component is still frequently observed and the attitude to overtransfuse patients is very difficult to correct, despite a number of guidelines for the use of blood and blood components have been published by various scientific societies and national health authorities. Australian benchmark studies demonstrated considerable variations of the Packed Red Cell (PRC) transfusion rate among institutions [1,2]. Other recent studies demonstrated overtransfusion and administration of double units as common errors [3,4]. Moreover, despite international guidelines are very clear about a sufficient dosage of plasma, underdosage of plasma seems to be an international phenomenon [4,5]. The reason for these observations remains unclear. Continuous information of Journal of Intensive and Critical Care ISSN 2471-8505 physicians involved in transfusion practice and monitoring has been claimed to increase the attention for transfusion practice and associated outcome. Since the complexity of a clinical situation such as in critical care might be a further explanation for the difficulty of guidelines adherence, actual reports about transfusion practice in critical care are wanted.
German transfusion guidelines (German transfusion guidelines [6] and Cross Sectional Guidelines for the Use of Blood and Blood Products [7] are very detailed. Critical illness is covered by sometimes very complex recommendations for the therapy with blood and blood products. They are based by a specific German transfusion law [8]; (First edition in 1998, last updates 2007 and 2009). For better adherence, guideline recommendations suggest a clinical transfusion corridor of situations in more or less detail for every blood product. This setting should provide sufficient support for accurate decisions in transfusion practice.
Despite these detailed support, the national German registry for transfusion errors ("IAKH transfusion error registry") contains a high percentage of errors with misuse (missing transfusion indication) (citation website of the registry report: http://www. iakh.de/auswertung2013.html), maladherence, or ignorance. The registry exists since 2009 and is fed from voluntary reports [9]. Although a collection bias with this hemovigilance clearly exists, the suspicion arises that in contrast to perfect preconditions transfusion practice among German physicians is poor for unknown reasons. For example, restrictive transfusion strategy is a repetitive error although the majority of physicians are educated about the TRICC trial [10]. Still, after more than a decade, implementation of this knowledge into German clinical practice seems to be very difficult.
To verify this thesis and to assess the underlying causes about the use of blood products, a retrospective analysis in an intensive care unit of a Level 1 university hospital was performed. Years after the publication of the detailed cross sectional guidelines, blood product use in accordance or disrespect of existing guidelines was recorded and evaluated for packed red cells, plasma and platelets. In case of non-alignment to guidelines, specific explanations were requested from ordering physicians.

Study population
A retrospective chart review was done in a 16-bed-perioperative Intensive Care Unit (ICU) at the University Hospital of Marburg, Germany. During a 12 month period, all data sets from transfused critical ill subjects either with red blood cells (PRCs), Fresh Frozen Plasma (FFP's) and Platelet Concentrates (PC's) thoroughly were reviewed or analyzed. Excluded from data analysis were patients on ExtraCorporal Membrane Oxygenation (ECMO) therapy or Intraaortic Balloon Pulsation (IABP) since their transfusion requirements are much higher and probably not addressed by guidelines. Our hypothesis was that blood product use in a German university ICU deviates from guidelines in a considerable percentage (>20%) of administrations.

Data collection, categorization and statistics
Transfused patients were identified by delivery and transfusion documents of the university blood bank. Documentation of transfusion indication was extracted from patient's charts. Laboratory and clinical data as well as other documentation were extracted both from paper and electronic medical records. Additionally, the responsible ward physician for the respective case was interviewed. For each blood product with unclear indication, the responsible physician's comment was requested and noted. Indications for each product was compared to given requirements in the cross sectional German guidelines (Querschnitts-Leitlinien (BÄK) zur Therapie mit Blutkomponenten und Plasmaderivaten). They were categorized as follows:

Packed red cells (PRC)
For one unit of red blood cells (PRC, 250-280 ml): 1) Transfusion with a unit of PRC is recommended below a hemoglobin content (Hb) ≤ 6 g/dl (EBM 1+); 2): Hb 6-8 g/dl only in patients with preexistent diseases that limit physiological mechanisms for anemia compensation (EBM 1+) or symptoms of anemia (tachycardia, hypotension, ischemic ST-changes, lactacidosis) (EBM 1+); 3): Hb 8-10 g/dl only with symptoms of anemia (tachycardia, hypotension, ischemic ST-changes, lactacidosis) (EBM 2c); 4): In patients with serious illness such as acute respiratory distress syndrome (ARDS), shock, sepsis, multiple trauma, massive blood loss postoperatively/gastrointestinal (GI) bleedings, ICU physicians considered a liberal transfusion strategy due to ongoing blood loss or active bleeding; 5): No transfusion is recommended above Hb 10 g/dl (EBM 1A). In stable patients, administration of double units without reassessment of the patient's clinical situation following reception of one unit is considered guideline discordant.

Fresh frozen plasma (FFP)
Transfusion of fresh frozen plasma (FFP) is recommended in following settings: 1) Microvascular bleeding following blood loss 100 ml/min; 2) red cell transfusion of more than 2 PRCs within 15 min; 3) transfusion of more than 4-6 PRC's; 4) If prothrombin time, aPTT and fibrinogen after prolonged blood loss, especially together with microvascular bleeding, are not available within brief; 5) PT ≤ 50%; 6) aPTT ≥ 45 s; 7) Fibrinogen ≤ 1 g/L; 8) indication is not to be found or setting is uncovered by guidelines.

Platelets (PC)
Chapter 2 of the cross sectional guidelines specifies the perioperative indications for platelets: 1) In cardiac surgery at increased postoperative bleeding in stable patients or below a platelet count of 20.000/μl; 2) in case of large and threatening bleedings for prophylaxis of a coagulopathy at <100,000 platelets/ μl; 3) during acute blood loss at <100,000 platelets/μl H g or of requirement of ≥ 1 PRC`s per day (WHO grade 3) 4) indication is not to be found or setting is uncovered by guidelines.
Data was analyzed following collection in an access database by Wilcoxon/Kruskall-Wallis ranked sum test. Significance was set at the alpha level of 5%.

Results
During the study period, 89 patients out of 650 ICU patients in total were transfused with blood products (13.7%). Treated patients were elderly (median 71 years) and originated from general, cardiac or orthopedic-trauma surgery departments. Most patients were admitted from other hospitals or departments critical care units (named "external" in Table 1). According to applied blood products, recipients were categorized in groups derived from the guidelines by the "Bundesärztekammer (German Medical Association) ( Table 2 -PRC, Table 3 -FFP, Table 4 -PC).

Transfusion triggers
Transfusions triggers and their approval by guideline indication can be seen in Table 5. Controlling for pre (p=0.2263) and post (p=0.0462) transfusion Hb content demonstrated guideline discomform PRC overtransfusion.
Guideline divergent plasma transfusion is not capable to change lab values for PT and aPTT (p<0.05).
The effect of platelet transfusion upon pre-transfusion platelet counts is associated to guideline conformity, indicating that a significant change could not be achieved by a not guidelineconform indication. The transfusion trigger "thrombocytopathy" increases the platelet threshold to 100 g/L.

Discussion
This is the first report about transfusion practice in critically ill patients in Germany. Although especially the German guideline recommendations are very detailed and are supported by variable strength of recommendation derived from evidence and grading, in approximately a third of investigated blood products guideline coverage could not be found. This result confirms aroused suspicions from the German IAKH error registry of blood products that the clinical use of blood has a high percentage of errors. The study might indicate a misuse of blood products doing ethical harm, as well as avoidable health care cost charging the health care society but also a potential harm to patients by unnecessary exposure to allogeneic blood. On the other hand, the study also demonstrates that complexity of the critical illness is a major explanation for liberal transfusion management and overtransfusion with all blood components.
The comparable degree of guideline discordance for all blood products indicates that the German system of transfusion practice still is not convinced by underlying evidence: For red cells, intensive care physicians still are using hemoglobin concentrations higher than 8 g/L as transfusion trigger, in a third of applications   Table 4 Indication category for platelets.
NOTE: For one request of platelets more than one indication could be chosen! I. In Cardiac Surgery at increased postoperative bleeding in stable patients or below a platelet count of 20.000/μl; II. In case of large and threatening bleedings for prophylaxis of a coagulopathy at <100,000 platelets/μl; III.
In acute blood loss at <100,000 platelets/μl or requirement ≥ 1 PRC`s per day (WHO grade 3) IV. No indication is given by the guidelines above 100 g/L. It seems that transfusion triggers given in guidelines are not specific enough for critically ill patients. For both plasma and platelets, establishment of point of care testing such as impedance aggregometry or PFA analysis might reduce transfusion rate, save cost and probably improves outcome [20]. In a multicenter study, the implementation of point of care testing reduced massive transfusion events by half, reduced plasma, red cell and platelet transfusion to respective maximal effects of reducing transfusion in 90% for FFP, 62% in RBC and 72% in platelets [21].
Considering the quality and accuracy of the transfusion guidelines of the German Medical Association, thirty percent of non-compliance seems to be quiet high. The search for an easy explanation requires the inclusion of various contributing factors. Napolitano and coworkers discuss the strength of underlying evidence and the level of acceptance of the recommending institution as critical factors for the speed and intensity for guideline implementation [22]. Thus, low evidence levels or doubts addressing the methodology underlying the restrictive transfusion recommendation are a reason for malcompliance among physicians as recently addressed [23][24][25]. However, especially recommendations with high evidence such as the prophylactic use of plasma were not adhered to in this study or other settings by a higher degree. In cirrhosis, prophylactic transfusion in the absence of planned procedure is widely (61%) used although not covered by evidence [26]. Furthermore, the complexity of clinical situations for critically ill patients might be contributing. In a Dutch study of general practitioners treating multimorbidity, the integration of patient-specific factors in medical decisions and the aim for patient-centered solutions was found to be interfering with applicability of guidelines [27]. However, in anticipation of the complexity of critical illness and its impact on transfusion behavior in this study, we already included transfusion categories for every blood product that were not exactly addressed in the guidelines. Although it might in stable patients without documented clinical symptoms. This is approximately consistent with "historical" studies but in contrast to existing guidelines. A clear reason remains speculative. For the double unit transfusion, a more procedural explanation arises -in more than half of cases (57.43%) a double as opposed to a single unit of PRC was demanded and administered. Consequently over transfusion could be demonstrated in this study by exceeding post transfusion Hb content. For this, multiple obstacles to the recommended single unit procedure seem to be responsible -it doubles work, delays the time to reach the target hemoglobin level and requires more attention by adding the additional step after the first unit. If the ordering physician is not aware of the risk of each unit of blood, the process is shortcut and simplified by a double unit order and administration. If this common practice can be avoided, the overall transfusion rate decreases and acquisition cost for blood is reduced [12]. A successful method to promote single-unit orders was achieved by special software for blood orders by either requesting a guideline covered indication (uncontrolled bleeding, instability) or restrict the order mode to single units. Within three subsequent observation years, it was able to decrease transfused units per day and double-unit transfusion by more than two-fold. [12].
In almost two thirds of applications (69.5%, average dose 7.27 ml/kg), plasma was transfused in insufficient doses [18]. One possible reason was the low sensitivity of routine laboratory parameters to indicate a coagulation disorder and the waiting period for the result. Harmful or ineffective plasma overuse [13][14][15] and prophylactic use [16][17][18] in critical ill patients could be overcome by coagulation concentrates [19].
Similarly, a third of platelet administrations were not in congruence with guideline recommendations due to assumed dysfunction. Platelet function measurement is not established ubiquously and therefore guidelines refer to number counts. Almost 70% of platelet administrations in this ICU were motivated by dysfunction, being performed at platelet counts  not compensate for the whole extent of malperformance, the results of this retrospective analysis do not support a major impact of critical illness upon the lack of guideline adherence.
Limitations of this study contributing to obtained results need further consideration. Many indications of blood products might be missed since this study was retrospective. However, the chart review was performed thoroughly and in close cooperation with the same ICU physicians in charge. Also, documentation of transfusion needs and success in the chart is common practice and part of guideline adherence. If no documentation was found, any other given reason for transfusion in accordance to guidelines was checked by blood losses and laboratory values. However, if documentation of blood loss equals clinical symptoms of ischemia in stringency and accuracy is debatable. Although troponin levels and EKG tracings, neurology consultations were reviewed, some suspicions of ischemia triggered the decision to transfuse (according to the interviews with the ICU physicians). Therefore, ischemia triggered RBC transfusion might have remained undocumented in some cases. German physicians are held to prove and document anemia symptoms, as well they are instructed to quit the habit of prophylactic blood use. In this study, especially for FFP and PC, the documentation of transfusion indication was missing in a considerable part of administrations. The explanation by ICU physicians that many clinical situations are not covered by guidelines seems too deliberate. Especially for platelets, an extensive number of clinical scenarios are covered by German guidelines.
Another issue should be considered: If we selected an institution with a low compliance to guideline adherence and transfusion medicine, the results would not be representative for German institutions. In the opposite, we selected a high quality and renowned teaching university level one trauma center and ICU with a very active and clinicaly involved transfusion medicine department. In consequence, the results obtained in this study can be considered representative, although performance and guideline adherence might be higher in a few other institutions in Germany, and even lower in many others.
However, what measure will improve guideline adherence? Actual medical information and Continuous Medical Education (CME) is best organized in university hospitals with large teaching facilities such as in our center. The impact of CME is poorly defined in relation to clinical outcomes, and efforts to standardize definitions of clinical outcomes are in varied stages of development [28]. Thus, there is no clear evidence that mandatory information and education as well as didactic education and passive dissemination strategies are effective in successful guideline implementation strategies [29]. According to authors of a recent assessment, effective implementation strategies included multifaceted interventions, interactive education and clinical reminder systems. The latter in the field of transfusion of blood and blood products might be software based solutions such as a monitoring and feedback program [30] or software enforcing guideline adherence [31][32][33]. Also, an intensive workshop for a week, a "transfusion camp", was very successful for transfer of transfusion knowledge to clinicians [34].
In conclusion, practice based educational together with profound electronic support might be the necessary method to improve guideline adherence in Germany.

Conclusion
In conclusion, transfusion guidelines were not followed by critical care physicians of a German university center in more than a third of applications. Double red cell unit administrations, under dosed plasma transfusion and platelet overuse together with a liberal attitude towards blood transfusion for critical ill were evaluated in this retrospective study. One reason for poor compliance might be the complexity of critical illness. Intensive education together with electronic support might be the most promising methods to improve compliance to guidelines.